Economic Load Demand Response in Energy Management Information Systems: Data Mining Methods in the Safety and Quality of Building Systems

Size: px

Start display at page:

Download "Economic Load Demand Response in Energy Management Information Systems: Data Mining Methods in the Safety and Quality of Building Systems"

Angel Fitzgerald
6 years ago
Views:

03 IEEE 0th Iteratioal Coferece o e-busiess Egieerig Ecoomic Load Demad Respose i Eergy Maagemet Iformatio Systems: Data Miig Methods i the Safety ad Quality of Buildig Systems Chig-Lug Li Dept.

1 03 IEEE 0th Iteratioal Coferece o e-busiess Egieerig Ecoomic Load Demad Respose i Eergy Maagemet Iformatio Systems: Data Miig Methods i the Safety ad Quality of Buildig Systems Chig-Lug Li Dept. of Electrical Egieerig Mig-Hsi Uiversity of Sciece ad Techology Hsi-Chu, Taiwa cll@must.edu.tw Hsueh-Hsie Chag* Dept. of Electroic Egieerig Ji-We Uiversity of Sciece ad Techology ew Taipei, Taiwa sschag@just.edu.tw Abstract Coordiatig ecoomic load demad respose (ELDR) strategy with eergy efficiecy ad iformatio techology of e-busiess multiplies the reductio i electricity usage. The real power, reactive power, ad total harmoic distortios of voltage ad curret of steady-state power sigatures cotai plety of iformatio eeded for detectig state trasitio ad agig of loads. O the other had, adoptig the values of power quality directly has the drawbacks of takig a loger time ad much memory for the data miig system of eergy maagemet iformatio system (EMIS). To effectively reduce the umber of power sigatures represetig load state trasitio ad agig sigals, a cluster classificatio techique of the power sigatures i the data miig system with differet mea ad stadard deviatios calculated by the Helliger distace are preseted i this paper. The high success rates of detectig state trasitio ad agig of loads from the proposed method have bee proved via experimets to be feasible i load operatios of EMIS applicatios. Keywords- eergy maagemet iformatio system; data miig; state trasitio; load agig; ecoomic load demad respose I. ITRODUCTIO Today it is ormal for buildigs to collect huge amouts of real-time eergy use data from eergy moitorig systems, icludig Programmable Logic Cotrollers (PLCs), Supervisory Cotrol ad Data Acquisitio Systems (SCADAs), etc. However, uless this captured data is shared ad aalyzed i a regular ad efficiet way to idetify problems ad provide solutios, this mass of data is merely overloaded with iformatio. A Eergy Maagemet Iformatio System (EMIS) is a performace maagemet system that eables idividuals ad orgaizatios to pla, make decisios ad take effective actio to maage eergy use ad costs. A EMIS ca be characterized by its deliverables, features, elemets ad support. Deliverables iclude the early detectio of poor performace, support for decisio makig ad effective eergy reportig. Features of a EMIS iclude the storage of data i a usable format, the calculatio of effective targets * Correspodig Author: H. H. Chag is with the Departmet of Electroic Egieerig, Ji-We Uiversity of Sciece ad Techology, HsiTie Dist., ew Taipei City, 3, TAIWA ( sschag@just.edu.tw). for eergy use, ad compariso of actual cosumptio with these targets. Elemets iclude sesors, eergy meters, hardware ad software. Essetial support icludes maagemet commitmet, the allocatio of resposibility, procedures, traiig, resources ad regular audits []. Fig. shows a typical eergy maagemet iformatio system i a buildig system. Sigle-phase two-wire ad sigle-phase three-wire systems are geerally used to ru a typical residetial or office buildig system. Ecoomic load demad respose (ELDR) programs use moitorig ad cotrol devices to maage what busiess takes from the power grid durig expesive high demad ad high eergy use periods. With ELDR, busiesses ca save moey by usig less eergy or shiftig electric load durig high demad ad high cost periods. Over the past two decades, there have bee may techiques proposed o ELDR for establishig various eergy cosumptio maagemets of the EMIS []-[9]. Ama et al. provide a comprehesive review of the literatures for the EMIS [0]. I additio, they preset a survey of the state of the art i eergy maagemet (EM) systems, applicatios, ad frameworks. Hou et al. propose a data miig method, combiig a rough set approach ad a artificial eural etwork (A), to detect ad diagose sesor faults based o the past ruig performace i the heatig, vetilatig, ad air coditioig (HVAC) systems []. Timbus et al. itroduce the commuicatio stadards Iteratioal Electrotechical Commissio (IEC) , IEC 6850, ad their extesios of distributed geerators (DGs) for active maagemet of distributed eergy resources (DER) to optimize the schedulig of the productio of wid eergy ad combied heat ad power (CHP) systems while maagig capacity costraits ad miimizig the total operatioal cost [4]. Byu et al. propose a smart eergy distributio ad maagemet system (SEDMS) to moitor iformatio about power cosumptio, a user s situatio ad surroudigs as well as cotrols appliaces usig dyamic patters based itelliget services ad adaptive demad maagemet to make the distributio ad cosumer more eergy efficiet ad itelliget. Accordigly, the results show that a reductio of the service respose time ad the power cosumptio are approximately 45.6% ad 9-7% respectively [6]. Silva et al. preset a ovel data miig framework for the exploratio ad extractio of actioable kowledge from data of eergy cosumptio aalysis geerated by electricity meters. I additio, the /3 $ IEEE DOI 0.09/ICEBE

2 proposed framework ca icorporate fuctioality for iterim summarizatio ad icremetal aalysis usig itelliget techiques [7]. Cha et al. itroduce a sigal processig commuity cocept of smart grids o the problems of price/load forecastig ad demad respose maagemet (DRM) ad optimizatio [8]. This paper presets a ovel evet idetificatio of ecoomic demad respose system cosistig of two models as show i Fig.. First, the power sigatures of power quality proposig techiques ad the stadardizatio of power sigatures are utilized to extract the eergy distributio features of the idividual evet state trasitio ad load agig at differet times. Secod, the Helliger distace is employed to classify load states accordig to the detailed eergy distributio. By usig the data miig model of Helliger distace, the umber of sigal features ca be reduced without losig its property. Takig the data collected from EMIS ad aalyzig the data usig data miig techiques are ot very ew. However, the researches of evet state trasitio ad load agig are valuable ad iovative ideas for ELDR of the safety ad quality of itelliget buildigs i EMIS. The accurate rate ad computatio requiremet of the proposed method were measured ad implemeted by real case studies. Accordigly, these results showed that the proposed method ca aalyze the sigals efficietly, thus ehace the performace of the EMIS ad efficietly maage eergy use ad costs. The ELDR program allows ed-users to get paid for reducig electricity demad durig high-priced hours. Participats receive the same price as geerators ad also realize savigs o their electricity bills. O the oe had, the ELDR algorithm proposed by the authors uses cluster aalysis to detect the evet state trasitio ad load agig i the EMIS, ad o the other the proposed method ca combie with iformatio techology (IT) to employ i e- busiess egieerig. For examples, decisio support of itelliget buildigs for e-busiess, applicatios of predictive load modelig to e-busiess, real-time ELDR aalytics for e-busiess, gree busiess, data ad kowledge maagemet for e-busiess, ad etc. Accordigly, the ELDR program proposed by the authors ca extesively idetify the real-time state chage of eergy market ad moitor busiess model agig for other service egieerig. This paper is orgaized as follows. The research motivatio, research purposes, ad relative papers reviews are addressed ad described i Sectio. The proposed methods of the paper is itroduced i Sectio, which icludes Helliger distace, stadardizatio, ad power sigatures of steady state; for examples, real power, reactive power, ad the total harmoic distortios of voltage ad curret. Based o the EMIS ad proposed data miig techiques, a series of ELDR aalyses for detectig evets state trasitio ad loads agig are implemeted for real case studies i Sectio 3. The discussios are addressed i Sectio 4 to explai the results of experimets. Fially, the coclusio ad further works are stated i Sectio 5. Fig.. Eergy maagemet iformatio system i a buildig system Fig.. Block diagram of a load evet idetificatio system II. PROPOSED METHODS The block diagram i Fig. icludes the algorithms of power sigatures of power quality, i.e. real power (P), reactive power (Q), total harmoic distortios of voltage (V THD ), ad total harmoic distortios of curret (I THD ), the processes of the stadardizatio techique, extractig the features from statistics of power sigatures, data miig system based o the Helliger distace, the idetificatio of load evet states, i.e. state trasitio or load agig, ad load demads moitorig ad load evets alarm. 77

3 A. Power Sigatures The curret ad voltage cosumed for a periodically oliear load ca be represeted by a Fourier series expasio. The appropriate coefficiets correspodig to the curret ad voltage i each harmoic are extracted from the results. The umber of terms represeted by the expasio determies the dimesio of the feature vector. The real power ad reactive power ca be computed by P P V0I 0 VI cos( V ) 0 () I ad Q Q VI si( V ) () I where is the harmoic umber; V ad 0 I are the average 0 voltage ad average curret, respectively; V ad I are the effective th harmoic compoets of the voltage ad curret; V ad I represet the th harmoic compoets of the voltage ad curret phase agles, respectively. The total harmoic distortios of voltage ad curret ca be computed by (3) ad (4), respectively. V I THD THD V V3 V4... % 00% (3) V I I3 I4... % 00% (4) I B. Stadardizatio Data expressios ca be made more efficiet if certai preprocessig steps are performed o the data miig system. Before miig, it is ofte useful to scale the power sigature iputs so that they always fall withi a specified rage. The approach for scalig iputs is to ormalize the mea ad stadard deviatio of the traiig set, ormalizig the iputs so that they will have zero mea ad uity stadard deviatio. These ca be computed by P ( P ) / (5) p Q Q p Q ( Q ) / (6) V / THD ( VTHD VTHD) VTHD (7) I / THD ( ITHD ITHD) ITHD (8) where the matrices P, Q, V THD, ad I THD are the origial power sigature iputs of real power, reactive power, ad the total harmoic distortios of voltage ad curret, respectively, the matrices P, Q, V THD, ad Q THD represet the ormalized power sigature iputs of real power, reactive power, ad the total harmoic distortios of voltage ad curret, respectively. The vectors ad cotai the mea ad stadard deviatios of the origial power sigature iputs. C. Helliger Distaces The squared Helliger distace is used to quatify the similarity betwee two ormal probability distributios i probability ad statistics. The Helliger distace is expressed i terms of the Helliger itegral, which was first developed i the 909 by Erst Helliger [], []. The ormal distributio is a cotiuous probability distributio. The squared Helliger distace ca be expressed as ( i j ) 4 i j i j, J e I i H (9) j Here, two ormal distributios I ad J are defied as follows: I ~ ( i, i ), J ~ ( j, j ) (0) where the parameter i this formula is the mea of the distributio. The parameter is its stadard deviatio; its. A radom variable with a Gaussia distributio is said to be ormally distributed ad is called a ormal deviate. The Helliger distace is closely related to the total variatio distace; for example, both distaces defie the same topology of the space of probability measures, but it has several techical advatages derived from properties of ier products. Helliger distaces are very well suited for the study of similarity of two ormal probability distributios. I this paper, the squared Helliger distace is employed to mie load evets i.e. load state trasitio ad load agig. III. EXPERIMETAL RESULTS A. Case Study Eviromet I case studies, the EMIS moitors voltage ad curret waveforms i a oe-phase electrical service etry powerig a collectio of loads represetative of the major loads i a commercial buildig. The data miig algorithm i the EMIS idetifies two loads with steady-state sigatures operatig o a 0-V commo bus. These loads iclude a electric rice cooker ad a refrigerator. Experimetal datasets were geerated by preprocessig the data o the voltage ad curret waveform of the load. Each fial sample cosists of 80 data samples ad 96 data samples obtaied over a period of 40 mi for a electric rice cooker ad a period of oe day for a refrigerator, respectively. A data miig simulatio program was desiged usig MATLAB. The program was ru to idetify load o a ASUS PC with a 3.3GHz Itel Core i3-0 CPU. Fig. 3 shows steady-state power sigature waveforms of a electric rice cooker. Each sample is a mea value of the real power ad reactive power over a period of 30 s ad a istataeous value of the V THD ad I THD at 30 s. The values of these steady-state sigatures are ot statioary whe the electric rice cooker is used for a log time because of 78

4 operatio temperature [3]. I additio, there is a state trasitio of power sigatures from rice steamig to keepig it warm. Fig. 5. State trasitio of a electric rice cooker Fig. 3. Power sigature waveforms of a electric rice cooker Fig. 4 shows steady-state power sigature waveforms of a refrigerator. Each sample is a mea value of the real power ad reactive power over a period of 5 mi ad a istataeous value of the V THD ad I THD at 5 mi. The values of these steady-state sigatures also chage whe the refrigerator is operated for a log time. Moreover, there is a state trasitio of power sigatures durig compressor operatio. Fig. 5 ad Fig. 6 show the state trasitio of a electric rice cooker ad a refrigerator for the power sigatures of P ad Q, respectively. The clusters of state trasitio are clear; however the load is the electric rice cooker or refrigerator. The variety of power of a refrigerator is bigger tha that of a electric rice cooker whe the compressor of refrigerator is workig or without workig. Table shows that the results of state trasitio for a electric rice cooker. The values of mea, stadard deviatio, ad Helliger distace are calculated for differet power sigatures whe a electric rice cooker steams ad keeps it warm. I the results, the state trasitio of a electric rice cooker ca be idetified by the proposed method for steadystate power sigatures. Table shows that the results of state trasitio for a refrigerator. The values of mea, stadard deviatio, ad Helliger distace are also calculated for differet power sigatures whe a refrigerator works with compressor ad without compressor. I the results, the state trasitio of a refrigerator ca be idetified by the proposed method for real power ad reactive power. Fig. 4. Power sigature waveforms of a refrigerator B. Results ()Case study, State Trasitio: I case study, the EMIS moitors state trasitio of a electric rice cooker ad a refrigerator i a oe-phase electrical service etry powerig a collectio of loads represetative of the major loads i a commercial buildig. Fig. 6. State trasitio of a refrigerator 79

5 TABLE I. THE RESULTS OF STATE TRASITIO FOR A ELECTRIC RICE COOKER I CASE STUDY Steam Warm Helliger Distace TABLE II. THE RESULTS OF STATE TRASITIO FOR A REFRIGERATOR I CASE STUDY Without Compressor With Compressor Helliger Distace ()Case study, Load Agig: I case study, the EMIS moitors load agig of a ew electric rice cooker ad a old electric rice cooker made i 998 i a oe-phase electrical service etry powerig a collectio of loads represetative of the major loads i the laboratory. Fig. 7 ad Fig. 8 show the load agig of the old electric rice cooker whe it works durig steamig ad warmig for the power sigatures of P ad Q, respectively. The clusters of load agig are very clear; however the load works durig steamig ad warmig. However, the power cosumptios of the old electric rice cooker are bigger tha that of the ew oe regardless of state. Fig. 8. Load agig of a electric rice cooker durig warmig Table 3 shows that the results of load agig for the electric rice cooker durig steamig. The values of mea, stadard deviatio, ad Helliger distace are calculated for differet power sigatures. I the results, the reactive power of the electric rice cooker is very small because the electric rice cooker is a resistace load. Accordigly, the load agig of the old electric rice cooker ca be oly idetified by the proposed method for real power of steady-state power sigatures. Table 4 shows that the results of load agig for the electric rice cooker durig warmig. The values of mea, stadard deviatio, ad Helliger distace are also calculated for differet power sigatures. I the results, the load agig of the old electric rice cooker ca be idetified by the proposed method for P, V THD, ad I THD of steadystate power sigatures. TABLE III. THE RESULTS OF LOAD AGIG FOR A ELECTRIC RICE COOKER DURIG STEAMIG I CASE STUDY ew Old Helliger Distace TABLE IV. THE RESULTS OF LOAD AGIG FOR A ELECTRIC RICE COOKER DURIG WARMIG I CASE STUDY Fig. 7. Load agig of a electric rice cooker durig steamig ew Old Helliger Distace

6 IV. DISCUSSIO I this paper, the electric rice cooker ad the refrigerator were chose as the experimetal samples to verify the performace of the EMIS. The loads, the electric rice cooker ad the refrigerator, are ofte basic electric appliaces for every day i the home. However, the represetatives of two samples have strog relevace with the research topic of the state trasitio ad load agig for the smart buildig. A appliace may have umerous load represetatios ad a load may be determied by various physical compoets. For example, a electric rice cooker has oly oe load, a heater, but has differet physical compoets for steamig ad warmig. A refrigerator has two loads, a fa motor ad a compressor, have differet physical compoets for cold storage ad freezig. For the state trasitio i case study, the state trasitio of the appliaces ca be idetified by the proposed method for steady-state power sigatures of P ad Q. The state trasitio of electric appliaces is irrelevat to harmoics of the load accordig to (3) ad (4); steady-state power sigatures of V THD ad I THD are difficult to idetify the state trasitio by the proposed method. For the load agig i case study, the load agig of the appliaces ca be idetified by the proposed method for oly steady-state P. It is ivestigated that the load agig is relevat to iteral resistace of the electric appliace from the result. From the results of case study, the iteral resistace of the electric appliace will be decreased ad the real power will be icreased because of load agig. However, it is very dagerous to icrease persistetly real power of the electric appliace because of load agig. Accordigly, a fire accidet could be happeed whe the real power is more tha rated real power of the electric appliace for a log time. V. COCLUSIO AD FUTURE WORKS The results of aalysis for EMIS ca idetify state trasitio ad load agig of itelliget buildigs ad to kow the coditio of use for loads. The users of buildig ca be remided to esure safety of appliaces by these results. Besides, some related policies of savig eergy, reducig CO, health ad safety care for hidde elderly ad the efficiecy of electric appliaces ca be established ad plaed by these results of smart buildigs. Based o experimetal results of EMIS, the real power ad reactive power of steady-state power sigatures for ELDR i EMIS ca be applied extesively to detect state trasitio ad agig of loads i the smart buildigs. The data miig system of proposed method are useful ad simple tools for makig decisios of detectig load agig i EMIS for itelliget buildigs. I the future works, the data miig techiques of proposed method ca be embedded i the o-itrusive eergy moitorig ad load idetificatio system to itegrate with EMIS. The EMIS will be a powerful system to maager eergy iformatio. ACKOWLEDGMET The authors would like to thak the atioal Sciece Coucil of the Republic of Chia, Taiwa, for fiacially supportig this research uder Cotract o. SC 0-- E REFERECES [] J. H. Hooke, B. J. Ladry, ad D. Hart, M.A., Eergy maagemet iformatio systems: Achievig improved eergy efficiecy: A hadbook for maagers, egieers ad operatioal staff, Office of Eergy Efficiecy of atural Resources Caada, 004. [] Z. Hou, Z. Lia, Y. Yao, ad X. Yua, "Data miig based sesor fault diagosis ad validatio for buildig air coditioig system," Eergy Coversio ad Maagemet, vol. 47, pp , 006. [3] S. V. Verdu, M. O. Garcia, C. Seabre, A. Gabaldo, ad F. J. Garcia Fraco, "Classificatio, filterig, ad idetificatio of electrical customer load patters through the use of self-orgaizig maps," IEEE Tras. Power Sys., vol., o. 4, pp , ov [4] A. Timbus, M. Larsso, ad C. Yue, "Active maagemet of distributed eergy resources usig stadardized commuicatios ad moder iformatio techologies," IEEE Tras. Idustrial Electro., vol. 56, o. 0, pp , Oct [5] M. A. Lisovich, D. K. Mulliga, ad S. B. Wicker, "Iferrig persoal iformatio from demad-respose systems," IEEE Security ad Privacy, pp. 0, Ja./ Feb. 00. [6] J. Byu, I. Hog, B. Kag, ad S. Park, "A smart eergy distributio ad maagemet system for reewable eergy distributio ad cotext-aware services based o user patters ad load forecastig," IEEE Tras. Comsum. Electro., vol. 57, o., pp , May 0. [7] D. D. Silva, X. Yu, D. Alahakoo, ad G. Holmes, "A data miig framework for electricity cosumptio aalysis from meter data," IEEE Tras. Idustrial Imformatics, vol. 7, o. 3, pp , Aug. 0. [8] S. C. Cha, K. M. Tsui, H. C. Wu, Y. Hou, Y. C. Wu, ad F. F. Wu, "Load price forecastig ad maagig demad respose for smart grids: methodologies ad challeges," IEEE Sigal Processig Magzie, pp , Sept. 0. [9] L. Heradez, C. Baladro, J. Lioret, D. Chiarro, J. J. Gomez-Saz, ad D. Cook, "A multi-aget system architecture for smart grid maagemet ad forecastig of eergy demad i virtual power plats," IEEE Commuicatios Magzie, pp. 06 3, Ja. 03. [0] S. Ama, Y. Simmha, ad V. K. Prasaa, "Eergy maagemet systems: state of the art ad emergig treds," IEEE Commuicatios Magzie, pp. 4 9, Ja. 03. [] E. Helliger, "eue begrüdug der theorie quadratischer forme vo uedlichviele veräderliche", Joural für die reie ud agewadte Mathematik (i Germa), vol. 36, pp. 0-7, 909. [] M. S. ikuli, Helliger distace, Ecyclopedia of Mathematics, Hazewikel, Michiel, Spriger, 00. [3] H. H. Chag, "o-itrusive demad moitorig ad load idetificatio for eergy maagemet systems based o trasiet feature aalyses," Eergies, vol. 5, o., pp , 0. 8

Radar emitter recognition method based on AdaBoost and decision tree Tang Xiaojing1, a, Chen Weigao1 and Zhu Weigang1 1

Radar emitter recognition method based on AdaBoost and decision tree Tang Xiaojing1, a, Chen Weigao1 and Zhu Weigang1 1 Advaces i Egieerig Research, volume 8 d Iteratioal Coferece o Automatio, Mechaical Cotrol ad Computatioal Egieerig (AMCCE 7) Radar emitter recogitio method based o AdaBoost ad decisio tree Tag Xiaojig,